Research On Optimization Of Dynamic Migration Of Virtual Machines For Load Balancing

As a popular network application mode, cloud computing can provide the flexible and storage and computing resources on-demand to the user through the network,which gradually becoming a hot academic research. Among them,the virtualization technology is an important support technology to promote the development of cloud computing, and it can realize resource sharing,simulation and isolation through virtualization, and it also improve the utilization rate of resources. However, The live migration of virtual machines is the key technology of virtual maintenance platform for reliable and stable operation.As an open source virtual machine software, many scholars carry out a variety of related research on the Xen. In this paper, it proposes an optimization strategy of machine live migration based on the analysis and research on the Xen migration mechanism, the main contents are as follows.Firstly, in this paper, due to the problem of load balancing of system resources in cloud computing environment, it proposes two solutions: fist, in order to solve the problem of migration destination node load metric does not fully consider, it extends the original four to six load metrics that CPU, memory, broadband, load changes, topology distance and hard drive capacity; second, This paper proposes the objective weight and the combination of the subjective weighting fusion weight strategy based on node load metric, which effectively solves the node load migration metric does not fully consider caused by separately considered one of the factors.Secondly, under the condition of the high rate of dirty pages, the traditional memory pre-copy algorithm will lead to a large number of dirty pages repeat transmission and a long transfer time. Therefore, this paper introduces the Optimized Grey model based on the traditional pre-copy mechanism, while using concurrency mechanism to resolve the issue of its timeliness generated in an iterative process.According to the historical access operation records of memory page, it uses the grey prediction model to predict the probability of the next round of memory pages become dirty. It predicts the dirty page rate of memory pages of virtual machines live migration, and it only transfer the memory pages which has the lower forecast dirty page rate. It can reduces the high rate of dirty pages of memory pages frequency heavy and the amount of data transmission, reduce the total migration time, improve the memory migration performance.Finally, under different environmental organizations experiment, this paper separately test the two algorithm, the destination node positioning strategy based on the combined weight and memory optimization algorithm based on gray model. Then, it collects and analyzes experimental data to verify performance of virtual machine migration, such as the migration time, downtime and load balancing, the experimental results show that the algorithm to ensure load balancing while also reducing migration time, improve migration efficiency.